You are on page 1of 5

Biosignal Based Human-Machine Interface for Robotic Arm

Saravanan N Madras Institute of Technology Mechatronics Department of Production Technology Chromepet,Chennai saravanan_n@ieee.org www.mitindia.edu Abstract
Human-robot interaction poses great challenges to disabled people who must be able to control the robot via a limited physical ability. A novel semi-autonomous humanrobot system where the robot requires only the commands from the human to perform its tasks as proposed. In this system a relatively more flexible and robust communication technique generates commands from Electromyogram (EMG) signals. Such an autonomous mobile humanoid robot will be able to assist in a workshop environment and interact with a human. Three aspects to reach this target are considered. First, how the brain (embedded system technologies) of the exoskeleton robot is constructed to be able to manipulate objects like humans do. Second, exploration of appropriate biosignals that can be used as an actuating quantity to move the Robotic Arm; and finally how programming of manipulation tasks are realized.

Mr.Mehboob Kazi M.S. Electronic Engineering Corporation Prof.G.Muralidharan Madras Institute of Technology

Human hand is attached with EMG sensor and the raw EMG signal is acquired. Since the amplitude of EMG signal is very small, we amplify it using a Instrumentation Amplifier and remove the noise using Twin-T Notch and Band Pass filters. After obtaining the signal, signal analysis is done through pattern classification algorithm such as Discrete Wavelet Transform. Analog signals are converted to digital form using ADC/Comparator which is then fed to a microcontroller which actuates the six servo motors of a robotic arm. Various works regarding electromyography involves various Human-Machine interaction in which blinking of eye could be used as an input for a mouse cursor and also includes studies of various hand gestures.

1. Introduction
This project describes a human-machine interface using Electromyogram (EMG) signal to artificially control the limb movement usually called Functional Electrical Stimulation and rehabilitation. Each muscle fiber has a potential and Motor Unit Action Potential generated by contraction of muscle is studied and corresponding actuation is provided to robotic arm. Examples of such applications are in air flight control and various rehabilitation aids.

2.Placement of EMG Electrodes
EMG Surface Electrodes are used in bipolar configuration where two electrodes with a small distance between each other are placed in the muscle to pick up the local signs with in the muscle of interest.A differential amplifier am[plifies the signals picked up from the two electrodes with respect to the signal picked up by a reference electrode.Because the interference signals from a distant source are essentially equal in magnitude and phase as detected by the two electrodes,the common mode rejection capability of the differential amplifier eliminates the unwanted signals

1 f .C are chosen such that the time period T=2 seconds.59M For this project we have fixed value of R6 =500 & R7=R8=50K. The first two amplifiers are working in the non-inverting mode but their inverting terminals are not grounded.signal analysis is done through Labview 7. 5. 1 2 * pi * r 2 * c2 1 1Hz= 2 * pi * r 2 * 0.INSTRUMENTATION AMPLIFIER Instrumentation Amplifier used for this project is LM324.C2. R11=R12=50K and R9=R10=10K. The feedback loops are connected with the inverting terminals. the signal is then passed through a band pass filter and the prominent frequencies between 10 Hz and 5 KHz is filtered out.e.1 using a Pattern Classification algorithm such as Discrete Wavelet Transform.SIGNAL ANALYSIS After obtaining the signal.3. The Instrumentation Amplifier is designed to have high input impedance.The values of resistors and capacitors are exactly identical to ensure high CMRR. After the 50 Hz signal is attenuated.1 *10 * * − 6 * R1 C1=0. let us assume the value of C2=0.) the 50 Hz signal is first filtered out using a Notch Filter. The third operational amplifier will act as a Differential Amplifier.R3 and R4.1*10^ −6 F1= We get R2=1.The values of R.C3 are provided to any DC offset.The Instrumentation Amplifier consists of three identical operational amplifiers. .1%. The high input impedance of 10 MegaOhm is provided by resistors R2. All the resistors are Metal Film military grade resistors with the tolerance level of 0.First. Therefore.1 farads 2KHz= R1=796 (820 ) The design of High Pass Filter is as follows.FILTERS The inherent noise of the electrical mains (i. a=100 and b=5 Gain= (1+2a) b= (1+200)5 =1005 4. The design of low pass filter is as follows: Fhigh=2KHz = 1 2 * pi * r1 * c1 1 2 * pi * 0. Wide Band Pass Filter The wide band pass filter is a Butterworth filter. It contains a high pass filter followed by a low pass filter.The capacitors C1.f1=1Hz for the high pass filter.All the capacitors are Tantalum capacitors with the tolerance of 1%.

This system seeks to enhance three kinds of muscle abilities: cooperation among several muscles. In the experiments. The sampling frequency for controlling the arm part and hand and wrist part were 60 and 100 Hz.55 and E2=2. 1 Flexor Carpi Radialis. For this purpose. Here we make use of LM329.(P) pronation. The figure shows the motion photos and EMG signals. considering the results of our research. Subjects A-D were fully functional. We then performed experiments on manipulator control using the EMG signals. 6 Brachialis). the Gaussian components. Rehabilitation training is beneficial before manipulation when the amputee’s muscle force declines with long lapses of time after amputation. The discrimination suspension and the online learning thresholds were determined by trial and error. The LLGMN structure was determined based on the number of electrodes.Experiments We conducted experiments with the developed manipulator system on eight subjects. Gaussian components were used to approximate the pdf of the sample data. (UF) ulnar flexion. and muscular contraction. . which were adequate for achieving high-discrimination performance A Discrimination Ability of the Hand and Wrist Motions First. respectively. If the subject was an amputee. In the experiments. we used six electrodes (: ch. ch. 5 Biceps Brachii. ch. 4 Supinator. from 21. and (HG) hand grasp) for about 30 s.to 31-year-old men and E-H were four men whose forearms were amputated in their early ages. we used the discrimination suspension rule and the online learning method. I first performed control experiments on the hand and wrist part. (HO) hand open.0. we placed four electrodes (ch. and examined the effect of online learning while the subject controlled this part for a couple of hours.1–4) on the muscles near the amputated part. ch. Subject E was trained using an EMG-based training system for prosthetic control. and two electrodes on the upper arm muscles (ch. 5 Biceps Brachii. 3 Pronator Teres.There were N=160(eight motions. we tried to control the joint angles based on the joint impedance model of the human forearm. ch. The numbers of components and learning samples were specified as one and 20 for each motion.6 Triceps Brachii). ch.(E) extension. and determined the thresholds as E4=0. 7 COMPARATOR The emg signals obtained are converted to digital form using Comparator. we examined the EMG pattern discrimination ability. ch. and the desired motions. to improve the feeling of control in the hand and wrist part. Finally. 20 for each motion) learning data inputs. (S) supination. The subject performed eight motions(K=8. 2 Flexor Carpi Ulnaris. timing of EMG generation. (RF) radial flexion. (F) flexion.Figure 6: Study of Hand Gestures 6.

ROBOTIC ARM The specifications and sketch of the robotic arm is as depicted in Table –I and figure 7 respectively.MICROCONTROLLER Here we make use of 89c51 microcontroller. The output of microcontroller is fed to a driver circuit made of DS 1669 9. Figure 7: Sketch of the Robotic Arm Figure 8: Link model of a manipulator .Table-I Specifications (a) Robotic Arm (b)Prosthetic Hand 8.

A Wavelet Tour of Signal Processing.”Intention Detection using a Neuro-Fuzzy EMG Classifier” [18] J." In: Proc of International [2] Weston J. 264:545-557. [21] Luo Zhizeng. 341-344. 1996.IEEE transaction on EMG.”Biomedical Technology and Devices Handbook”.George Zouridakis. 2000." Machine Learning. "The norms and variances of the Gabor." PhD Thesis. [6] Ciaccio E.China.Michael Norris.59-65 [12] Bronzino.”Using Singular Eigen values of Wavelet Coefficient as the input of SVM to recognize motion patterns of the hand” . Cristianii N. [8]Cortes C. 1995.Dario Farina. HONG kong.”Biomedical Engineering Handbook”.CRC press.Roy. "The Nature of Statistical Learning Theory.Kilby.Hongzhou Institute of Electronic Engineering. References [1] Wang Rencheng.Robot Research Institute. [20] Zhizeng Luo. Shawe-Taylor J.Wiley Edition.Pages 170182." In: Advances in Neural Information Processing Systems 12.”Multi-Pattern Recognition of the forearm movement based on SEMG”. "Discussion on Various Methods of EMG Processing for the Control of Prostheses. [11] Khandpur R S. [5] Mallat S." New York:Springer-Verlag. “Electromyography: Detection.1995.Pages 178-187 [13] David Prutchi. Beijing [10] Platt J. "Multiresolution Representation and Wavelets. "Biosignal pattern recognition and interpretation system. Beijing: China Machine Press. "Multi-class support vector machines. PA.Electromyography Department. Philadelphia.Gao Jian. 1993.Electromyography-Unlocking the secrets of back pain. Dunn S. (In Chinese) [4] Boltezar Simonovshi M. et al. Massachusetts: MIT Press.L. [3] Stephane Mallat. “Handbook of Biomedical Instrumentation.Newzealand [19] Mark S. Akay M. 2003." IEEE Eng in MedandBio Mag.Hussein and Malcolm H.20: 273-297." Journal of Sound and Vibration. "Large margin DAGs for multiclass classification.Pages 1-305 [14] James Moore. Roberto Merletti.Emley. "Support Vector Network.Tata McGraw Hill. (6):106 [7] Vapnik V N. 1988. University ofPennsylvania.Auckland University of Technology.10." Royal Conference on Biomedical Engineering.Semmlow. [9] Vapnik V N.Granat. The Nature of Statistical Learning Theory.Xiaoliang Ren and Yutao Jia.Gholam Hosseini. “Wavelet Analysis of Surface Electromyography Signals”. Processing and Applications” [16] John L. Watkins C.Second Edition”.”Biosignal and Biomedical Image Processing-Matlab based Applications”.Donald Gilmore and Serge H. Vapnik V. Moriet and general harmonic wavelet functions.CRC Press [15] Marco Pozzo.”Design and Development of Medical Electronic Instrumentation”.Pages 177-241 [17] Sherif E.H.